# First Cal National Rankings from Sagarin

USA Today's Jeff Sagarin publishes football team rankings based on purely numeric data, which sounds reasonable on the surface, but the devil is in the details.

For starters - when each season starts - his system has very little data to base an analysis upon. As the season progresses, Sagarin's rankings are computed from thousands of games where the results are used to compute a team strength rating. Using statistical methods, his software determines the rating for each team that "best" explains the wins and losses to all the teams it has played.

But that "best" needs explanation. A best data fit does not mean a perfect fit - it merely means that no other number exists that fits the overall data better than the one computed - even if it is inaccurate - and even if the computed rating does not accord with all of the actual wins and losses. To illustrate, it is entirely possible that one team's rating might exceed that of another, even though the team lost a game to the other. Ultimately, the ratings can be thought of as a sort of average of all the games played during a season, with all the ups and downs blended in.

And it's still not that simple, because politics intrude into this mathematical world in the form of the BCS prohibition against using the margin of victory in the calculation. They say a win must be counted as just a win - regardless of whether the margin of victory was 30 points or 3. There is a sound motive behind this, to make it unproductive for teams to run up scores to get better ratings to improve the changes of being selected to a more prestigious bowl game.

Because of this prohibition, Sagarin publishes three sets of rankings, one that complies with the BSC requirements, one that includes the mathematical effects of the winning margin, and a thrid that aveages the first two. Because a margin of victory does supply meaningful information, Sagarin calls the second set of rankings his "Best Predictor" - because those numbers have been proven to be the best at predicting the outcome of future games. (Sagarin's "Best Predictor" is used at the BearPlay game at this web site. Click Here to see the game.)

Oddly, the BCS strictures have actually improved the quality of Sagarin's Best Predictor ratings. By making score run-ups less useful, actual margins of victory have become a somewhat better measure of the relative strengths of teams - and hence result in better predictive ability.

In data released today, Sagarin's BCS-compliant rankings have Cal 2nd in the country, probably because of a) Cal's win over a strong team - an event that most others in the top-10 have not yet experienced, and b) the paucity of data with only one game played so far.

Sagarin's Best Predictor - the one that includes winning margins - has Cal at 11th nationally, an assessment more likely to accord with coach's and sports writer's polls when those are announced.

The following tables presents Tuesday's Sagarin Ratings for the top ten teams with one game played.

 Team Average Rank BCS Rank Best Rank LSU 93.99 1 91.16 3 97.39 2 Southern California 90.10 2 91.59 1 88.32 10 Georgia Tech 90.09 3 87.88 9 92.32 3 California 89.91 4 91.44 2 88.10 11 Ohio State 89.84 5 89.64 5 89.57 5 Oklahoma 89.73 6 84.84 17 98.05 1 Florida 88.37 7 85.91 14 90.94 4 Wisconsin 88.36 8 88.67 7 87.58 13 Georgia 88.22 9 88.72 6 87.26 15 West Virginia 87.21 10 85.34 15 88.90 8

(Cal fans - note with pleasure Sagarin's use of "Southern California" to identify the Figeroa Drive Behemoth, a label the Trojans loath. Way to go, Jeff)